applog-listen tsx units keep stale in-memory code after source push

2026-10-06 measured: Venus listener scripts were clean at commit 6964dbe while applog-listen@pluto and @venus still ran pre-change 8404168 code loaded 2026-09-22. A support restart at 15:02:54 UTC gave Pluto new PID 2965047 and loaded dual-rail null normalization; push alone was not live. Scope every configured applog-listen unit on Venus: @enamel, @mars, @pluto, @terra, @venus (active in systemctl at filing); @ayudarg exists in maintainer registry/systemd but is currently inactive (MainPID 0), so classify its intended state rather than blindly start it. Source owner is Venus scripts/applog-listen.ts and scripts/lib/burst-classify.ts; per-unit maintainer ownership varies. Add an explicit restart-and-readback step after relevant script/lib pushes OR a version probe/alert comparing running image with desired source version. Acceptance: force red with clean on-disk newer code but old PID still running, then show the gate reports stale or the restart updates each intended-active unit; verify PID/start, loaded revision, cursor continuity/catch-up, and heartbeat without assuming a git push restarts tsx. Do not change Pluto migration/apply under this row.